# Ben: chronic anxiety — the evidence base and the treatment ranking

Compiled 27 August 2026 from a four-thread literature review. This file exists because the supplement-and-hormone framing that dominated earlier sessions was the wrong frame, and several confident claims made in those sessions were wrong.

## The presentation

Lifelong, by his own account 10+ years and probably longer. Background tension rather than panic. Not snappy, not impairing: he functions well and describes himself as high functioning. Present on days he takes nothing. Pregabalin 75 mg relieves it cleanly without mellowing him.

That is a textbook generalised anxiety picture, chronic course. **Bruce et al. 2005 (Harvard/Brown, 12-year prospective) confirms GAD runs a chronic course with low recovery and high recurrence.** Framing the goal as management rather than cure is the evidence-consistent framing. https://pubmed.ncbi.nlm.nih.gov/15930067/

**The single most important structural fact: he has never had one adequate trial of any first-line treatment for it.** Ten years of supplements and no course of anything. Everything below follows from that.

**He is a low-dose pregabalin responder — corrected 27 August 2026.** An earlier framing here made much of 75 mg being half the lowest dose ever tested in a GAD trial. That population fact is true and it is **not decisive for him**: he has tried 150 mg and reports it as almost too much, taking him into a loose, ultra-relaxed state that is pleasant socially but past where he wants to be for work. So 75 mg is not a bandaid dose in his case, it is his dose. Do not push him toward the studied range on the strength of the trial data alone. The dose-response curve is a population average and he sits well left of it.

**What started the search:** a single phenibut trial a couple of years ago. He describes it as an eye-opener not for the experience but because it was the first time he registered how tense his normal state is. He had no prior reference point. Everything since — the supplements, the compounds, this whole line of enquiry — dates from that. Clinically this is the most useful thing in his history: he has felt the target state once, chemically, which means the MCT question is not "what would relaxed even be" but "can he reach that without a compound."

**Phenibut itself:** GABA-B agonist, unscheduled in most places, and carrying a dependence and withdrawal profile considerably worse than pregabalin's — withdrawal can run to severe agitation, insomnia, psychosis and seizures, and it is documented at therapeutic-range doses. He used it once and is not using it now. Given he describes an ongoing search for deep relaxation, state plainly and once that this is the one compound in the space not to return to. Do not lecture; he has not asked to use it.

## Corrections — claims made in earlier sessions that the evidence does not support

| Claim made | Status |
|---|---|
| "Tolerance to pregabalin's anxiolytic effect develops faster than tolerance to its side effects" | **RETRACTED.** Backwards on both halves. Tolerance to sedation and dizziness is documented within weeks (Owen 2007). Anxiolytic tolerance has never been demonstrated out to 12 months. Feltner 2008 held 450 mg for 8+24 weeks with relapse curves that did not converge; the 2025 meta-analysis found the drug-placebo gap **grew** over time (MD −1.23 at 2 wk → −3.31 at 6–12 mo). BAP's formulation is the defensible one: the relative potential for tolerance "is not established". |
| "Low dose propranolol is the mechanistically precise answer for his somatic tension" | **RETRACTED as a recommendation.** Steenen 2016: 8 RCTs, **202 participants total, zero in GAD** — "insufficient to support the routine use of propranolol in the treatment of any of the anxiety disorders." The 2024/25 update agrees. The "somatic anxiety only" teaching traces to Tyrer & Lader 1974, **n=6 per subgroup, one week.** Mechanism-led hypothesis, not an evidence-led one. |
| "Pregnenolone is causing his tension" | **RETRACTED.** He falsified it directly: tense on days he takes none, tense for a decade before he started it. |
| "Pregnenolone feeds the aromatisation cascade and raises E2" | **RETRACTED.** No human study shows oral pregnenolone raising testosterone or estradiol in men. Sripada 2013 (n=31 men, 400 mg): pregnenolone ×3.2, allopregnanolone ×7.1, **DHEA-S DOWN ~5%**, DHEA and cortisol unchanged. It does not behave as a "mother hormone" feeding the androgen axis. https://pmc.ncbi.nlm.nih.gov/articles/PMC3648625/ |
| "His weak 5AR means the calming allopregnanolone arm is quiet" | **RETRACTED.** Measured human PK says allopregnanolone rises **7-fold** on oral pregnenolone. The elegant DHT-ratio story was wrong. |
| "Elevated E2 drives his tension/mood" | **RETRACTED for anxiety specifically.** See below. The E2 concern is real but it is a prostate and cardiovascular concern, not a mood one. |

**One correction that stands:** pregnenolone lowering DHEA-S by ~5% matters, because his DHEA-S was already low at 103.4 against a floor of 139.7, and he was put on DHEA 25 mg to fix it. Pregnenolone is working mildly against that, not for it.

## Estradiol and anxiety in men: the link is not there

This matters because it stops a wrong argument being used to push a right action.

- **The headline citation for "high E2 causes mood symptoms in men"** is LIFE-Adult (n=3,925). Trace the units: the difference is **26.2 vs 23.0 pg/mL**. A 3 pg/mL gap deep in the normal range, cross-sectional, **with no anxiety measure at all**. https://pubmed.ncbi.nlm.nih.gov/29107881/
- **No study has ever tested whether E2 at 80–90 pg/mL causes anxiety in men.**
- **Counter-evidence:** trans women routinely run E2 at 100–200 pg/mL; an 18-month prospective cohort (n=178, HADS) found depression fell and **anxiety did not worsen**. https://pubmed.ncbi.nlm.nih.gov/32777129/ PATCH/STAMPEDE men on transdermal estradiol had **better** quality of life than men on LHRH agonists.
- **"Crashed E2 causes anxiety"** has **zero** controlled human data. Estrogen deficiency in men demonstrably causes hot flashes, fat gain and bone loss with a threshold around **10 pg/mL** (Finkelstein). Mood has never been tested in either direction.
- **"Optimal E2 is 20–30 pg/mL for mood"** has no primary source. It is clinic and forum material.

**Also relevant: supraphysiological testosterone is probably not driving his tension either.** Bhasin 2001 took 61 men to **2,370 ng/dL for 20 weeks** — "sexual function, visual-spatial cognition and mood did not change significantly at any dose." https://pubmed.ncbi.nlm.nih.gov/11701431/ Tricker/Bhasin 1996, 600 mg/wk for 10 weeks, no change in anger including by **observer** report from a spouse or parent. https://pubmed.ncbi.nlm.nih.gov/8855834/ What is real is a **4–12% idiosyncratic hypomania/irritability responder rate** at ≥500–600 mg/wk that nobody can predict from baseline — but he is well below that dose and has a decade-long tension history that predates any of it.

**So: keep pushing for the E2 draw, but push for it on the PSA and cardiovascular grounds, which are his own data (PSA 0.505 → 1.146 → 0.864). Do not claim it will fix the anxiety.**

## The treatment ranking

Ordered by effect size × evidence quality × drive preservation, which is his binding constraint.

| Option | Effect size | Evidence | Drive cost |
|---|---|---|---|
| **Pregabalin at a real dose** | HAM-A MD **2.79**, d **0.50** | Very high. #2 of 22 drugs in Slee (Lancet, 25,441 pts), placebo-level acceptability | Low. Sedation, not anhedonia |
| **Metacognitive therapy** | **g 1.61** in GAD; g 0.69 **over CBT** | Moderate. Small trials, allegiance effect live | None |
| **CBT** | g 1.01 vs placebo in GAD | Very high | None |
| **Agomelatine 25–50 mg** | HAM-A MD 3.28 | Moderate, Servier-dominated | Lowest of any antidepressant |
| **HRV biofeedback** | g 0.83 between-group | Moderate. Small unblinded trials | None |
| **Buspirone 45–60 mg** (adjunct) | d 0.17 mono; class SMD 0.58 | Moderate, old trials | None. **Reverses** SSRI sexual dysfunction |
| SSRI/SNRI | HAM-A MD 2.4–3.2 | Very high | **Real. 40–60% blunting, ~59% sexual dysfunction** |
| Quetiapine | HAM-A MD 3.60, largest of any drug | High | High. D2 antagonism, metabolic |
| Beta blockers, CBD, vortioxetine | Null to trivial | Low | n/a |
| Ketamine, psilocybin, MDMA **for GAD** | **No completed GAD RCT exists** | None | n/a |

### Why pregabalin tops it for him specifically

1. **It already works on him, cleanly, by his own report.** That is the strongest single datapoint available and it beats any population statistic.
2. **CACNA2D1 rs17584499 CC** — normal binding site at the drug's actual target. Genotype match.
3. **His anxiety is glutamate-driven and GABA-responsive** (GABRA2 TT), which is the mechanism α2δ ligands act on. Pregabalin reduces presynaptic glutamate release.
4. **It is EU-licensed for GAD**, first-line in the Canadian and WFSBP-2012 guidelines, alternative-first-line in BAP.
5. **It avoids the two things that would cost him most.** No D2 antagonism (his dopaminergic tone is already high: MAOA TT, DRD2 rs6277 AA), and no serotonin reuptake inhibition (TPH2 TT + HTR1A CC means SSRIs work uphill, with a poor efficacy-to-blunting ratio).
6. Onset **in days**, not weeks. Separated from placebo by **day 4** in Kasper 2009, beating venlafaxine.
7. It **increases** slow-wave sleep, where benzodiazepines reduce it — relevant to the 3–5am waking.

### The dose problem, which is the headline finding

**75 mg/day is half the lowest dose ever tested in a GAD trial, and there is no efficacy evidence for it at all.**

- Licensed range: **150–600 mg/day**, divided BID or TID
- **150 mg failed to separate from placebo** in Feltner 2003 and was the weakest arm in the pooled dose-response (Lydiard 2010, n=1,854)
- Dose-response **plateaus at 300 mg/day**; peak anxiolysis around 450 mg
- BAP: "efficacy... is more marked at higher daily doses (200 mg or higher)"

He has been taking a bandaid dose of a drug licensed for his condition, intermittently, and concluding the category does not work.

### The costs, stated honestly — these are real and they matter for him

- **Weight gain: ~20% of patients over a year** (Montgomery 2013: 19.4% in the GAD cohort). He is cutting toward 10% body fat. This is a direct conflict.
- **Sexual dysfunction. The SmPC lists erectile dysfunction AND decreased libido as *Common* (≥1/100).** One uncontrolled series (Hamed 2018, n=75) found 41.3% sexual dysfunction, not dose-related, reversible on stopping. BAP's honest verdict: the incidence "is uncertain" — **no GAD trial ever measured it with a validated instrument against an SSRI comparator.** For a dating coach on TRT this is the single most important trade to name.
- **Cognition.** Memory impairment, disturbance in attention and confusion are all *Common* on the SmPC. Far better than benzodiazepines on objective psychomotor testing (Hindmarch 2005) but not inert.
- **Dependence.** Real but concentrated: **1.6% misuse in the general population vs 3–68% among people who abuse opioids** (Evoy 2017). Bonnet & Scherbaum's 106-study review found only **four individuals** meeting dependence criteria without prior substance history, and rated the risk **lower than other sedatives**. Taper over at least a week; longer is better.
- Schedule V in the US, Class C / Schedule 3 in the UK.

### PRN versus daily — the honest answer

**Entirely off-evidence.** Every GAD RCT used scheduled BID/TID dosing. No guideline addresses intermittent use. Pharmacologically a single dose does produce acute anxiolysis (a single 200 mg dose beat placebo in a human approach/avoidance paradigm), so it is not implausible the way PRN sertraline would be. But t½ is ~6.3 h, so PRN means repeated full washouts, and effect-contingent intermittent dosing is the exposure pattern most associated with the drift from prescribed to self-administered use.

## What the other options are for

- **Metacognitive therapy has the largest effect size in this entire file (g 1.61 in GAD)** and is the best-fitting psychological approach for a high performer, because it targets beliefs *about* worry — that it is uncontrollable, and simultaneously useful and protective — rather than challenging worry content. Content-level CBT tends to get resisted by people who experience worry as instrumental to performance. 6–12 sessions. Evidence caveat: mostly Wells' own group, allegiance effect live. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2018.02211/full
- **Agomelatine** is the fallback if pregabalin's weight and sexual costs are unacceptable: no sexual dysfunction signal, lowest emotional blunting of any antidepressant (28% vs 60% escitalopram), no discontinuation syndrome, no weight gain. Costs: LFT monitoring schedule, one sponsor behind nearly all the evidence, and it **failed its non-inferiority margin** against escitalopram in the one honest head-to-head. Not available in the US; available across Asia.
- **Buspirone** is a poor monotherapy (d 0.17, weakest active class) and a genuinely underused adjunct at **45–60 mg/day** — the standard 10–15 mg is subtherapeutic and is why it has a reputation for not working. Landén 1999: added to an SSRI it improved sexual function in **58% vs 30% placebo**.
- **HRV biofeedback (g 0.83)** is the best-supported non-drug intervention aimed at the somatic layer specifically. Resonance-frequency breathing, ~6 breaths/min, 10–20 min daily, deliverable on consumer wearables. Not a supplement, which is the category he has correctly written off.

## What to rule out, and why

- **SSRIs/SNRIs as first choice.** Not on efficacy grounds — they work — but on fit. TPH2 TT plus HTR1A CC predicts working uphill, and the blunting risk is the specific thing he is trying to avoid. Langley 2023 showed escitalopram reduces reinforcement sensitivity in **healthy** volunteers, so the blunting is the drug, not the illness. Reasonable second-line if pregabalin fails.
- **Quetiapine.** Largest HAM-A effect of any GAD drug and correctly relegated. D2 antagonism against an already-high dopaminergic tone is exactly wrong for him.
- **Adaptogens and cortisol interventions.** FKBP5 CC, efficient HPA recovery, AM cortisol 10.3 mid-range. Whole category is a dead end.
- **Silexan.** If 80 mg/day of the actual preparation was run for 8–10 weeks with no effect, the herbal lane is properly closed.
- **Beta blockers, CBD, ketamine, psilocybin, MDMA for GAD.** No GAD evidence, in several cases no GAD trial has ever been run.
- **Guanfacine and clonidine.** Two negative adult RCTs for guanfacine (Neylan 2006, Davis 2008); clonidine's entire controlled anxiety evidence base is one n=23 crossover from 1981.

## The behavioural read

This is the same pattern as the rest of the file, one layer up. Inputs optimised relentlessly — compounds, doses, genotype theories — while the actual intervention was never run. The reason ten years of magnesium, theanine, lavender and 75 mg PRN produced nothing is not that his anxiety is exotic or treatment-resistant. It is that none of that constitutes a treatment.

When he asks for "something potent that fits my genetics", the honest answer is that the genotype-guided part is real but small, and the large part is that a common, treatable condition has gone untreated. Do not let the search for a bespoke mechanism substitute for running a standard adequate trial.

---

## The recommendation, settled 27 August 2026

**Sequence by cost, not by effect size. Run the zero-cost option first.**

1. **Keep pregabalin 75 mg PRN exactly as it is.** It works for him, at his frequency the dependence risk is negligible, and there is no reason to disturb something that is working. He was explicit that he finds it useful and never said otherwise; an earlier session mischaracterised him as dismissing the drug class, which he did not.
2. **Do metacognitive therapy first.** Largest effect size in the review (g 1.61 in GAD), zero cost to weight, libido, drive or dependence, and time-limited at 8 to 12 sessions.
3. **Cut caffeine timing.** He named caffeine and poor sleep as the two things that make it pronounced. CYP1A2 AC means slow clearance, so this is a duration problem, not a sensitivity one. Free lever, self-identified trigger.
4. **Hold daily pregabalin 150 to 300 mg in reserve.** It is a legitimate licensed option and the trade is real, not theoretical: ~20% weight gain over a year against a man cutting to 10% body fat, and an uncertain but non-trivial sexual dysfunction rate against a man whose profession and TRT protocol both make that costly. If MCT does not move it in three months, this is the next step and he would be taking that trade knowingly rather than by drift.

**Do not push daily pregabalin as the headline recommendation.** The earlier framing did, and it read as being sold a drug he had not asked for.

### Answering the addiction question precisely

Distinguish three things that get conflated:

- **Addiction/misuse.** At a few times a month at 75 mg, effectively nil. The whole dependence literature concerns daily 300 to 600 mg. General population misuse is 1.6%; among people who abuse opioids it is 3 to 68%. Bonnet & Scherbaum's 106-study review found **four individuals** total who met dependence criteria without a prior substance history.
- **Physical adaptation.** Real on daily use, requires a taper of at least a week. This is the same class of thing as stopping an SSRI or a beta blocker, not addiction.
- **Dependence on continuation.** The one that actually traps people. Stopping brings the underlying anxiety back fast: 50% of Feltner's responders relapsed within **23 days** of switching to placebo. That is relapse of the condition, not withdrawal, but it feels identical from the inside and is why people stay on.

### Metacognitive therapy: the concrete referral

- **MCT-Online (mct-online.org)** is run by the MCT Institute itself; all therapists are at the Institute's highest training level. English, plus several other languages. 45-minute sessions, 15-minute trial available. Suits Bangkok and Dubai time zones since most therapists are UK and Scandinavia based.
- **Adrian Wells takes patients there.** He originated MCT. Others at Level 2: Henrik Nordahl, Hans Nordahl, Sverre Urnes Johnson, Odin Hjemdal, Dave Bowman, Gerry McCarron.
- **Verify MCTI Level 1 or Level 2 registration.** Anyone can call themselves a metacognitive therapist; the Institute is explicit that improperly applied MCT can worsen symptoms.
- **What it targets:** two belief sets, not worry content. Negative ("my worry is uncontrollable and harmful") and **positive** ("worrying keeps me sharp, prepared, ahead of the market"). The positive set is the sticky one in high performers and is the reason content-level CBT gets resisted by them.
- **Techniques:** Attention Training Technique (structured auditory attention exercise, ~12 min daily), detached mindfulness, worry postponement, then direct behavioural testing of the beliefs.


## Is his anxiety genetic? Handle this carefully

Asked directly on 27 August 2026. Two separate answers, and conflating them is the trap.

**Is anxiety heritable?** Yes, meaningfully. GAD heritability runs roughly 30 to 40%. Family history is the informative signal, and it is now recorded.

**His family history, given 27 August 2026: positive on both sides, neither formally treated.** Father had some form of depression per his mother's account, no detail, never treated as far as he knows. Mother is clearly anxious and physically tense, socially variable, and is the parent **he recognises himself in** — the same quality of being on edge rather than settled. Uncles noticeably more relaxed. This is a far stronger heritability signal than anything in his SNP panel.

**Do not let him read that as a ceiling.** Heritable is not unmodifiable, and "it runs in my family" is exactly the shape of belief MCT dismantles. If he reaches for the family history as evidence the tension is fixed, that is the treatment target announcing itself, not a prognosis.

**Flag for the therapist and for future sessions:** he volunteered that he may not have a clear reference point for what relaxed feels like, and that he sees very few people who strike him as genuinely relaxed. Left unexamined that belief makes the tension look universal and therefore not worth treating, and it will also distort how he judges whether treatment is working.

**Do the psychiatric SNPs in his panel explain it?** No, and this is the weakest part of his whole genetic file. MAOA rs6323, TPH2 rs4570625, HTR1A rs6295, GABRA2 rs279858, COMT rs4680 are all **candidate-gene** associations, and psychiatric candidate-gene research largely failed replication. Border et al. 2019 (Am J Psychiatry) tested 18 leading candidate genes for depression in samples up to 620,000 and found no support for any of them. Anxiety is polygenic across thousands of variants of tiny effect; a consumer panel reading a handful of candidate SNPs is not a genetic explanation of anything.

**This applies to claims made in earlier sessions**, including several of mine: the "constitutionally tense phenotype" argument built on MAOA TT plus ADRB2 Arg/Arg plus GABRA2 TT, and the CACNA2D1 CC explanation for his pregabalin response, are all weaker than they were presented. They are plausible mechanism stories, not established findings. **The cardiovascular and APOE parts of his genetic file are a different tier of evidence entirely and should not be tarred with this.**

### Why the genetics must NOT go in his therapist briefing

He asked whether to include it. The answer is no, and the reason is clinical rather than tidiness.

**MCT's primary target is the belief that worry is uncontrollable.** A genetic self-explanation is the single most powerful support that belief can have: it says the tension is hardwired, permanent, and outside his agency. Walking into MCT holding a printout of anxiety SNPs actively arms the thing the treatment is trying to dismantle.

More broadly: be careful about handing him biological explanations for the anxiety at all. He is drawn to them, they are mostly not well supported, and they work against the intervention most likely to help him.